Compiler hangs during patmat analysis

[Christewart]

Reproduction steps

Scala version: 2.13.18

Christewart/bitcoin-s-core@ec55801

$ git clone git@github.com:Christewart/bitcoin-s-core.git
$ git checkout ec558010eb9d2cae59f4ac96fa8db429d540e156
$ sbt "coreJVM/compile"

And that should never finish compiling. I've waited 10+ minutes. I've observed this on both macOS/linux so doesn't seem OS specific. Compile times on current master ( d092aed967ac83c6c6265d5907bf01d3e2d36bbc ) ~25 seconds on apple silicon mac

Problem

I would expect scalac to succeed or fail in compiling the source files

[som-snytt]

A quick kill -3 would yield a stack trace which may or may not be helpful.

[Christewart]

@som-snytt Ive attached the result of kill -3 in the attached file

compiler-bug.txt

[lrytz]

It seems to be stuck in exhaustiveness analysis, so -Xno-patmat-analysis (or maybe -Xnon-strict-patmat-analysis) would help as a workaround. Can you pinpoint a specific commit that causes the issue to show up?

[Christewart]

It seems to be stuck in exhaustiveness analysis, so -Xno-patmat-analysis (or maybe -Xnon-strict-patmat-analysis) would help as a workaround. Can you pinpoint a specific commit that causes the issue to show up?

It's the commit I linked in the OP

[SethTisue]

It's the commit I linked in the OP

I'm able to reproduce the problem at that commit, but if I try to compile its parent commit, it doesn't compile.

What's the last previous commit where compilation succeeds? What we want is to see the diffs between the broken version and the last working version.

[Christewart]

It's the commit I linked in the OP

I'm able to reproduce the problem at that commit, but if I try to compile its parent commit, it doesn't compile.

What's the last previous commit where compilation succeeds? What we want is to see the diffs between the broken version and the last working version.

Christewart/bitcoin-s-core@d092aed

[SethTisue]

here's a single link showing the diffs: https://github.com/Christewart/bitcoin-s-core/compare/d092aed..ec558010

I see that core/src/main/scala/org/bitcoins/core/script/interpreter/ScriptInterpreter.scala has a large pattern match (inside private def loop) with many cases, where most of the cases were of the form e.g. case OP_IFDUP :: _ on a List, but changed to e.g. case OP_IFDUP +: _ on a Vector. that seems like a likely culprit, and a likely basis for constructing a standalone reproducer